The file has been created in such a way that the web browser is ignoring the non-html parts of the document, while the image renderer is ignoring the parts that

The file has been created in such a way that the web browser is ignoring the non-html parts of the document, while the image renderer is ignoring the parts that make up the html page. The first part probably isn't too hard, since most web browsers go to great lengths to render non-standard html in a sensible way, I'm not too sure about the second part. I'm guessing the jpeg spec has some variable length space in some kind of file header that the html for the page can be put in to. I read something similar a while back (I think it was called a Jafar attack) where a clever person worked out how to create a file that was both a valid .gif image and .jar java executable.